By using this site, you agree to our Privacy Policy and our Terms of Use. Close

Forums - Sony - Sony to use Nvidia Tegra for PSP2; explains delay

Jo21 on 14 July 2010
darkknightkryta said:

Weren't the rumors saying the PSP2 was going to use a PowerVR chipset?


yes they had a deal for next gen powervr chip   but i guess its not ready yet

but if psp2 will be based on GT6 it will support version 2-3 shaders



Around the Network
Soleron on 14 July 2010
jarrod said:
...

GameCube could do soft/self shadowing, normal mapping, specular highlights and subsurface scattering?  News to me. ;)

Agreed. The effects are better and the chip is easier to work with. However the number of textures/polygons look to me to be Gamecube level, in that I can imagine Mario Sunshine being at the limit of what the chip could display assuming the E3 demos show off the chip well.

The GPU core in 3DS is based on DMP's PICA chip, which dates back to 2006 as the first iteration btw.  I'm not sure why you're using ATi chips to try and give a timeline versus Tegra?

The Gamecube used an ATI chip that dates in architecture from 2001. Since I'm guessing the  Pica200 chip is about GC capability, and that Tegra is about GF6 capability, I'm measuring the differential in desktop-GPU-years.

Of course both chips were designed much later than those dates, but handheld GPU capability tends to be in step with desktop GPU capability, just X years apart from needing Y number of die shrinks to get the power envelope to ~1/30 of its previous size.

Restated: they can make a chip as capable as 9 years ago now but with 1/30 of the power use.

Sony also has a higher power and cost budget to work with, unless their strategy on pricing and battery life has changed since the PSP.





Soleron on 14 July 2010
Jo21 said:
darkknightkryta said:

Weren't the rumors saying the PSP2 was going to use a PowerVR chipset?


yes they had a deal for next gen powervr chip   but i guess its not ready yet


Well, that was a rumour and so is this. I happen to trust this website more than those that were reporting PowerVR, because of its track record on Nvidia. Take your pick.

Those sites reporting Sony were to use PowerVR were also reporting Nintendo were to use Nvidia. The fact it didn't happen undermines their credibility.



dahuman on 14 July 2010
Soleron said:
jarrod said:
...

GameCube could do soft/self shadowing, normal mapping, specular highlights and subsurface scattering?  News to me. ;)

Agreed. The effects are better and the chip is easier to work with. However the number of textures/polygons look to me to be Gamecube level, in that I can imagine Mario Sunshine being at the limit of what the chip could display assuming the E3 demos show off the chip well.

The GPU core in 3DS is based on DMP's PICA chip, which dates back to 2006 as the first iteration btw.  I'm not sure why you're using ATi chips to try and give a timeline versus Tegra?

The Gamecube used an ATI chip that dates in architecture from 2001. Since I'm guessing the  Pica200's chip is about GC capability, and that Tegra is about GF6 capability, I'm measuring the differential in desktop-GPU-years.

Of course both chips were designed much later than those dates, but handheld GPU capability tends to be in step with desktop GPU capability, just X years apart from needing Y number of die shrinks to get the power envelope to ~1/30 of its previous size.

Restated: they can make a chip as capable as 10 years ago now but with 1/30 of the power use.



he was just being a smart ass, the cube can do all those and more, but you have to work on a fixed pipeline if you want a lot of other things which can be very inefficient on performance. Pica unfortunately is in that same state but with a very updated design and supports GL ES1.1(which is 1.5 really) so people can do much more since it supports the common important effects of today's games.



Soleron on 14 July 2010
dahuman said:
...

he was just being a smart ass, the cube can do all those and more, but you have to work on a fixed pipeline if you want a lot of other things which can be very inefficient on performance. Pica unfortunately is in that same state but with a very updated design and supports GL ES1.1(which is 1.5 really) so people can do much more since it supports the common important effects of today's games.


The way I've seen it argued is that a fixed pipeline allows Nintendo to do a lot of effects using less power than shaders, at the cost of raw throughput. They of course had the option of a conventional GPU like PowerVR's and didn't choose it.

Since designers are making games with that chip in mind (and indeed, if Sony doesn't do something soon, the lead platform), they can make games look good even without the polygon power, and Nintendo gets to keep the DS's good battery life.



Around the Network
dahuman on 14 July 2010
Soleron said:
dahuman said:
...

he was just being a smart ass, the cube can do all those and more, but you have to work on a fixed pipeline if you want a lot of other things which can be very inefficient on performance. Pica unfortunately is in that same state but with a very updated design and supports GL ES1.1(which is 1.5 really) so people can do much more since it supports the common important effects of today's games.


The way I've seen it argued is that a fixed pipeline allows Nintendo to do a lot of effects using less power than shaders, at the cost of raw throughput. They of course had the option of a conventional GPU like PowerVR's and didn't choose it.

Since designers are making games with that chip in mind (and indeed, if Sony doesn't do something soon, the lead platform), they can make games look good even without the polygon power, and Nintendo gets to keep the DS's good battery life.

that's what I was thinking as well actually, it's all a trade off.



jarrod on 14 July 2010
Soleron said:
jarrod said:
...

GameCube could do soft/self shadowing, normal mapping, specular highlights and subsurface scattering?  News to me. ;)

Agreed. The effects are better and the chip is easier to work with. However the number of textures/polygons look to me to be Gamecube level, in that I can imagine Mario Sunshine being at the limit of what the chip could display assuming the E3 demos show off the chip well.

The GPU core in 3DS is based on DMP's PICA chip, which dates back to 2006 as the first iteration btw.  I'm not sure why you're using ATi chips to try and give a timeline versus Tegra?

The Gamecube used an ATI chip that dates in architecture from 2001. Since I'm guessing the  Pica200's chip is about GC capability, and that Tegra is about GF6 capability, I'm measuring the differential in desktop-GPU-years.

Of course both chips were designed much later than those dates, but handheld GPU capability tends to be in step with desktop GPU capability, just X years apart from needing Y number of die shrinks to get the power envelope to ~1/30 of its previous size.

Restated: they can make a chip as capable as 9 years ago now but with 1/30 of the power use.



In terms of geometry, I'd agree, GC-ish (6-12m/pps in actual game scenarios) is probably on the money for what we'll see.  It's worth keeping in mind though that for the 3D to work, 3DS functionally has to draw all those triangles, textures and effects twice... it's more equivalent to 2 GameCubes running in parallel in that sense.  If all that horsepower were put into a single image, it'd be well beyond any of the last gen consoles.

Another thing to note, Kojima said the models used in the Snake Eater 3D demo were actually closer to MGS4 level geometry than MGS3.  I think the platform might be a bit more robust than we have indications for so far... also worth keeping in mind all those 3rd party demos were thrown together in under 2 months, which makes things like Snake Eater 3D or Resident Evil Revelations all the more impressive.

Also, PICA200 has basically no real relation to Flipper, and the shader architecture seems pretty far off from TEV... honestly, that just seems like a pretty reaching comparison to make.



jarrod on 14 July 2010
dahuman said:
Soleron said:
jarrod said:
...

GameCube could do soft/self shadowing, normal mapping, specular highlights and subsurface scattering?  News to me. ;)

Agreed. The effects are better and the chip is easier to work with. However the number of textures/polygons look to me to be Gamecube level, in that I can imagine Mario Sunshine being at the limit of what the chip could display assuming the E3 demos show off the chip well.

The GPU core in 3DS is based on DMP's PICA chip, which dates back to 2006 as the first iteration btw.  I'm not sure why you're using ATi chips to try and give a timeline versus Tegra?

The Gamecube used an ATI chip that dates in architecture from 2001. Since I'm guessing the  Pica200's chip is about GC capability, and that Tegra is about GF6 capability, I'm measuring the differential in desktop-GPU-years.

Of course both chips were designed much later than those dates, but handheld GPU capability tends to be in step with desktop GPU capability, just X years apart from needing Y number of die shrinks to get the power envelope to ~1/30 of its previous size.

Restated: they can make a chip as capable as 10 years ago now but with 1/30 of the power use.



he was just being a smart ass, the cube can do all those and more, but you have to work on a fixed pipeline if you want a lot of other things which can be very inefficient on performance. Pica unfortunately is in that same state but with a very updated design and supports GL ES1.1(which is 1.5 really) so people can do much more since it supports the common important effects of today's games.

Wii could maybe, but it's my understanding that GameCube's nerfed memory architecture inhibited a lot of use for the TEV.  Of course, developer apathy seems to have made pushing TEV on Wii an equally futile proposition.  At least Wii managed normal mapping and rim lighting. :/

Also, the NV2A  in the Xbox was technically a fixed pipeline chip.  People tend to associate the fixed nature of Flipper as what held it back, but it was really just familiarity with the APIs.  That's something that Xbox and now 3DS don't suffer, thankfully.



Soleron on 14 July 2010
jarrod said:
Soleron said:
...



In terms of geometry, I'd agree, GC-ish (6-12m/pps in actual game scenarios) is probably on the money for what we'll see.  It's worth keeping in mind though that for the 3D to work, 3DS functionally has to draw all those triangles, textures and effects twice... it's more equivalent to 2 GameCubes running in parallel in that sense.  If all that horsepower were put into a single image, it'd be well beyond any of the last gen consoles.

I thought it did 50% of the pixels on the screen to each eye? So it's producing two images of half-size only.

Another thing to note, Kojima said the models used in the Snake Eater 3D demo were actually closer to MGS4 level geometry than MGS3.  I think the platform might be a bit more robust than we have indications for so far... also worth keeping in mind all those 3rd party demos were thrown together in under 2 months, which makes things like Snake Eater 3D or Resident Evil Revelations all the more impressive.

Yes. Then again, the GC did not really progress beyond Mario Sunshine and Metroid Prime, so maybe they'll never use the 'extra' power.

Also, PICA200 has basically no real relation to Flipper, and the shader architecture seems pretty far off from TEV... honestly, that just seems like a pretty reaching comparison to make.

Agreed. Tegra 2 will, however, be stronger than PICA, that's what I was getting at.





jarrod on 14 July 2010
Soleron said:
jarrod said:
Soleron said:
...



In terms of geometry, I'd agree, GC-ish (6-12m/pps in actual game scenarios) is probably on the money for what we'll see.  It's worth keeping in mind though that for the 3D to work, 3DS functionally has to draw all those triangles, textures and effects twice... it's more equivalent to 2 GameCubes running in parallel in that sense.  If all that horsepower were put into a single image, it'd be well beyond any of the last gen consoles.

I thought it did 50% of the pixels on the screen to each eye? So it's producing two images of half-size only.

Another thing to note, Kojima said the models used in the Snake Eater 3D demo were actually closer to MGS4 level geometry than MGS3.  I think the platform might be a bit more robust than we have indications for so far... also worth keeping in mind all those 3rd party demos were thrown together in under 2 months, which makes things like Snake Eater 3D or Resident Evil Revelations all the more impressive.

Yes. Then again, the GC did not really progress beyond Mario Sunshine and Metroid Prime, so maybe they'll never use the 'extra' power.

Also, PICA200 has basically no real relation to Flipper, and the shader architecture seems pretty far off from TEV... honestly, that just seems like a pretty reaching comparison to make.

Agreed. Tegra 2 will, however, be stronger than PICA, that's what I was getting at.



That's how the screen works, not the the GPU.  They render the scene twice, for each eye.

I'd also argue RE4 was pretty impressive on Gamecube, progression didn't exactly halt in 2002.